Self-gravity and quasi-stellar object discs

It is well known that the outer parts of QSO accretion disks are prone to selfgravity if heated solely by orbital dissipation. Such disks might be expected to form stars rather than accrete onto the black hole. The arguments leading to this conclusion are reviewed. Conversion of a part of the gas into high-mass stars or stellar-mass black holes, and the release of energy in these objects by fusion or accretion, may help to stabilize the remaining gas. If the disk extends beyond a parsec, however, more energy is probably required for stability than is available by turning half the gas into high-mass stars. Small black holes are perhaps marginally viable energy sources, with important implications (not pursued here) for the QSO spectral energy distribution, the metallicity of the gas, microlensing of QSO disks, and perhaps gravitational-wave searches. Other possible palliatives for selfgravity include accretion driven by nonviscous torques that allow near-sonic accretion speeds and hence lower surface densities for a given mass accretion rate. All such modes of accretion face major theoretical difficulties, and in any case merely postpone selfgravity. Alternatively, thin disks may not exist beyond a thousand Schwarzshild radii or so (0.01 parsec), in which case QSOs must be fueled by gas with small specific angular momentum.